COMMUNICATIONS
PRICE-LEVEL STABILITY, PRICE FLEXIBILITY,
AND FISHER’S BUSINESS CYCLE MODEL
Scott Sumner
Introduction
In a series of recent papers, De Long and Summers (1986, 1988)
argued that “increased wage and price flexibility can easily be destabilizing because of the Mundell effect.” They also suggested that
early studies ofthe Phillips Curve by Irving Fisher (1923, 1925) had
reached a similar conclusion. The National Industrial Recovery Act
of 1933 (NIRA) was cited as a possible example of how policies
designed to reduce price flexibility could have a stabilizing effect on
output.
One motivation for this paper is to show that De Long and Summers
appear to have misunderstood both the implications of Fisher’s business cycle model and the impact of the NIRA on the level of U.S.
output. De Long and Summers have apparently confused the concept
of price-level instability (i.e., erratic fluctuations in the price level)
with price flexibility (the absence ofconstraints separating the actual
price level from its Walrasian equilibrium value). Although Fisher
believed that price-level instability was the primary cause of output
fluctuations, he would not have agreed with the proposition that a
reduction in price flexibility is stabilizing.
A second purpose of this paper is to see how Fisher’s model performs in the 10 years following its original publication. Fisher (1925)
developed a model in which a distributed lag of inflation rates was
used to explain deviations of output from the trend during 19 15—22.
It will be shown that Fisher’s “Phillips Curve” model continues to
perform reasonably well at explaining output fluctuations until July
1933. After July ~933, output falls sharply below the level predicted
Cato Journal, Vol. 9, No. 3 (Winter 1990). Copyright © Cato Institute. All rights
reserved.
The author is Associate Professor of Economics at Bentley College.
719
CATO JOURNAL
by Fisher’s model. This structural break is associated with the implementation of the NIRA. The poor performance of Fisher’s model
after July 1933 may help explain why the monetary approach to
business cycles, which was dominant during the 1920s, was virtually
abandoned after the mid-1930s.
Fisher’s Model of the Business Cycle
Fisher saw the business cycle as largely a “dance of the dollar.”,
He diff~redfrom most modern researchers, however, by his assumption that the causality ran strictly from prices to output.’ In his model,
price fluctuations affected output because “When producers get
higher prices, they do not at first have to pay higher wages and
salaries” (Fisher 1925, p. 180). Another difference between Fisher’s
model and more recent models featuring sticky wages is that Fisher
did not assume rational (or even adaptive) expectations.2 During the
early 1930s, Fisher began to stress the impact of deflation on debt
burdens as an important factor in the rise in bankruptcies and
unemployment.
Figure 1 is reprinted from Fisher’s 1925 paper. It shows trendadjusted output(T) and a distributed lag of inflation rates (P’) extending over 114 months. The reported correlation is .941. (See the
Appendix for a detailed description of Figures 1 and 2.) Using a
similar technique, Fisher found somewhat lower price-output correlations during earlier periods of U.S. history.
Although Fisher was the first economist to provide a mathematical
representation of the Phillips Curve, the idea of a short-run tradeoff
between inflation and output was an essential partof macroeconomic
theory during the 1920s. A number of contemporary economists,
including Keynes (1923), Cassel (1922), Pigou (1927), Hawtrey
(1923), and Robertson (1922), saw price instability as a major cause
of the business cycle.3 This view led many of these economists to
recommend that monetary policy be directed toward the goal of
1Fisher (1925, p. ~86) rejects the idea ofestimating the correlation between changes in
current output and future inflation on the grounds that this “would involve the apparent
absurdity of bringing some of the effects earlier than their supposed cause.”
2Fisher was familiar with adaptive expectations, having used this concept in his work
on nominal interest rates. The fact that Fisher’s model performed as well as it did
during the interwar years may be due to the fact that the expected rate of inflation was
probably close to zero under the gold standard.
3Pigou (1927) estimated that approximately one-halfofthe output variability was caused
by price-level fluctuations.
720
PRICE-LEVEL STABILITY
FIGURE 1
TREND-ADJUSTED OUTPUT AND DISTRIBUTED LAG OF
INFLATION RATES OVER 114 MONTHS
NOYE: T represents the volume of trade or the trend-adjusted output; P’
represents a distributed lag of monthly inflation rates.
SOURCE: Fisher (1925, p. 188).
price-level stabilization.4 Wage and price controls were not regarded
as an effective substitute for monetary stability.
Afterthe mid-1930s there was a sharp decline in interest in proposals to stabilize the price level. To see why the neoclassical macroeconomics ofthe 1920s was eclipsed by Keynesian economics, one might
find it helpful to see how well Fisher’s model explained output
fluctuations in the period leading up to the 1936 publication of the
General Theory.
Figure 2 shows the result of updating Fisher’s model to the
1923—35 period. The correlation between the predicted and the
actual output series was only .256. This low correlation, however, is
somewhat misleading since, during the decade from January 1923
through July 1933, the correlation was .85 1. The variability of output
4Both Hawtrey and Cassel proposed the use of central hank cooperation under a gold
exchange standard, Keynes and Wicksell preferred a completely managed fiat currency.
Fisher’s “Compensated Dollar Plan” would have adjusted the price of gold inversely
to changes in the price level.
721
FIGURE2
(T) AND
(F’), 1923—35
DEvIATIONS OF INDUSTRIAL PRODUCTION FROM TREND
DISTRIBUTED LAG OF MONTHLY
T.30
INFLATION RATES
.075P’
.20
.050
.10
.025
0
0
—.025
—.20
—.050
—.30
—.075
—.40
—.100
.50
—.125
—
PRICE-LEVEL STABILITY
in that decade was somewhat higher relative to the variability of
inflation than was the case during 1915—22, but the qualitative sitnilarities between the two periods would suggest no apparent reason
for rejecting Fisher’s model.
The model also predicts a rapid recovery from the depression
during 1933. Industrial production did rise by almost 50 percent
during the first half of 1933—a much sharper rebound than occurred
during the early stages of recovery from the 1920—22 depression.
After July 1933, however, the economic recovery stalled and industrial production did not return to its July levels until August 1935.
One ofthe more unusual aspects of the U.S. economy during 1933
was the rapid increase in the price level during a time when output
was far below capacity. McCloskey and Zecher (1984) compared
weekly changes in the foreign exchange value of the dollar with
wholesale prices; they argued persuasively that the inflation that
occurred during the spring of 1933 was caused by a depreciation in
the dollar.
De Long and Summers suggest that the NIRA helped increase
prices during 1933. If so, then its major impact could not have occurred before June or July. The NIRA was passed on June 16 and
was first implemented on July 15. The impact of the NIRA codes
shows up most clearly in the Bureau of Labor Statistics’ aggregate
wage series (see Table 1). Note that the aggregate wage rate rose 22
percent (19 percent in real terms) between July and September 1933.
It seems clear that only governmental intervention in the labor market could have produced such a large increase in wages during a
period of 25 percent unemployment.
This sharp increase in real wages was associated with a dramatic
slowdown in industrial production. During the first half of 1933
industrial production had increased simultaneously with the price
level. After the adoption of the NIBA codes, however, industrial
production dropped sharply while prices continued to increase.
In retrospect, it is not surprising that De Long and Summers
viewed Fisher as a supporter of New Deal attempts at reducing price
flexibility. Fisher argued that, under most circumstances, monetary
policy should aim at price-level stabilization. After the sharp deflation of 1929—33, Fisher felt that some degree of reflation was necessary before prices were stabilized. In fact, Fisher was an enthusiastic
supporter of President Roosevelt’s policy of reflation.
As previously noted, however, the primary cause of rising prices
during 1933 appears to have been the devaluation of the dollar.
Unlike wage-price controls, this action was not inconsistent with
Fisher’s free-market ideology. Thus, although in 1933 Fisher clearly
723
CATO JOURNAL
TABLE 1
SELECTED ECONOMIC INDICATORS, MONTHLY, 1933
Industrial
Average
Exchange RateR
WPI”
Production~
Wagesd
May
3.361
3.422
3.433
3.579
3.932
61.0
59.8
60.2
60.4
62,7
30.8
30.7
29.0
31.0
36.2
.406
.404
.405
.401
.398
June
4.136
65.0
41.6
.394
July
August
September
October
November
December
4.650
4.503
4.665
4.668
5.150
5.116
68.9
69.5
70.8
71.2
71.1
70.8
45.3
43.5
41.2
38.9
36.7
37.0
.399
.459
.488
.493
.496
.503
Month
January
February
March
April
‘Exchange Rate = Dollar Price of Pounds, from the Federal Reserve Bulletin.
“WPI = Wholesale Price Index as measured by the Bureau of Labor Statistics, (1926
= 100).
~Industrial Production is taken from the Federal Reserve Bulletin, (1947—49
100).
riAverage Wages = Production-Worker Average Hourly Earnings (in dollars) from
Employment and Earnings.
favored reflation and then price-level stabilization, he also argued
that “the introduction ofcodes under the NRA, as well as the processing of taxes under the AAA, seemed to have a deterrent effect upon
the activities of industry” (Fisher 1934, p. 362). Fisher favored pricelevel stability but not price inflexibility.
Fisher employed a quantity theoretic framework to analyze
changes in the price level. While subsequent Keynesian theorists
argued that monetary policy was ineffective during the Great Depression, Fisher believed that the 1929—33 deflation could have been
averted by a more expansionary monetary policy. In fact, Fisher
(1933, p. 347) believed that such a policy would have been adopted
“had Governor Strong of the Federal Reserve Bank of New York
lived, or had his policies been embraced by other banks and the
Federal Reserve Board and pursued consistently after his death.”
Fisher was also aware that the constraints imposed by the international gold standard could at times prevent the monetary authority
from taking the steps necessary to stabilize the price level. He, there724
PRICE-LEVEL STABILITY
fore, proposed a “Compensated Dollar Plan” whereby the value of
the dollar would be adjusted each month in proportion to changes in
the price level (i.e., the price of gold would be adjusted inversely
with the price level). One of the reasons why Fisher was so supportive of Roosevelt’s policies was that the devaluation of 1933 in a
sense represented a belated implementation of the Compensated
Dollar Plan.
Conclusion
The view that price-level instability is an important source of
output fluctuations was a key element of pre-Keynesian macroeconomic theory. By using Fisher’s Phillips Curve model to predict
output fluctuations over the period from 1923 to 1935, we can gain
insights into the events that led to the Keynesian revolution. Fisher’s
model provides no explanation for the relatively low level of production from mid-1933 to mid-1935. There is circumstantial evidence
that the adoption of the NIRA in June 1933 helped abort a vigorous
economic recovery that was already under way.5
Although it appears that De Long and Summers were incorrect in
suggesting that the NIRA might have had a stabilizing impact, this
fact should not be interpreted as a rejection of their model. They
noted that on theoretical grounds the NIRA’s effect was ambiguous
and that “The NIRA’s encouragement of cartelization may have had
contractionary macroeconomic effects” (De Long and Summers
1986, p. 1043). Neither should the fact that the adoption of NIRA
codes was associated with a pause in the recovery be construed as
ruling out other contributing factors that may have delayed recovery,
such as trade wars or bank panics.
In a subsequent paper, De Long and Summers (1988) argued that
recovery occurred after the announcement of the NRA. As noted
earlier, the weekly data examined by McCloskey and Zecher show
a clear relationship between inflation and devaluation. With respect
to wages (a nontraded good), the impact of the NIRA does not occur
until after July.
Dc Long and Summers (1988) also stated that “In an accounting
sense all of the Great Depression was due to a collapse in monetary
velocity that cannot implausibly be traced to deflation; recall that
real broad money balances (M2) did not decline between 1929 and
1932” (p. 275). It is true that all four components of the quantity
equation declined by approximately one-third between 1929 and
5Weinstein (1980) reached a similar conclusion.
725
CATO JOURNAL
1933. But if prices had been artificially restrained from declining
after 1929, then, holding Federal Reserve policy constant, real money
6 Thus, their observation has
balances would have declined sharply.
no bearing on the stabilizing properties of price flexibility.
A more serious problem with their argument is that De Long and
Summers seem to repeatedly confuse price flexibility with price
instability. For instance, in the 1988 paper they cite evidence from
earlier studies showing that price shocks affect future output movements. This finding is consistent with numerous conventional macro
models, including new classical models where unperceived pricelevel changes affect output (with a lag caused by adjustment costs),
as well as “Keynesian” models featuring the sort of wage rigidities
described by Fischer (1977).~
Appendix
In his 1925 paper, Irving Fisher used a detrended, seasonally
adjusted index of output (T) that was computed by Warren M. Persons. The Wholesale Price Index published by the Bureau of Labor
Statistics was used to calculate the inflation rate. The monthly inflation rate was defined as the percentage change between the last
period’s price level and the next period’s price level, multiplied by
six. The distributed lag of inflation rates was derived by taking a
weighted average of 114 lags of the inflation rate. The distribution
ofweights is approximately normal when the time axis is transformed
logarithmically.
In updating Fisher’s model to the 1923—35 period, I used the
seasonally adjusted industrial production series computed by the
Board ofGovernors ofthe Federal Reserve System. The output series
shown in Figure 2 represents the percentage difference between
actual industrial production and its predicted value derived by
regressing monthly industrial production on time, over the 1913—46
period. Since the sample period ends in the middle of the Great
Depression, the series was detrended over a period including several
business cycles. I also attempted to use the same weights on lagged
inflation as in Fisher’s 1925 paper. However, as Fisher was somewhat
vague regarding his methodology, there may be some slight
differences.
6One could,of course, argue that had deflation not occurred, then the moneymultiplier,
and therefore M2, would not have declined. This argument would not apply, however,
to deflations accompanied by sharp drops in the monetary base, as occurred during
1920—2 1.
‘Although Fischer’s model is usually regarded as “Keynesian,” it is clearly much closer
to the pre-Keynesian models of Fisher and Pigou.
726
PRICE-LEVEL STABILITY
References
Cassel, Gustav. Money and the Foreign Exchange after 1914. New York:
Macmillan, 1922,
De Long, J. B., and Summers, L. H. “Is Increased Price Flexibility Stabilizing?” American Economic Review 76 (December 1986): 1031—44.
De Long, J. B., and Summers, L. H. “Is Increased Price Flexibility Stabilizing?: Reply.” American Economic Review 78 (March 1988): 273—76.
Fischer, Stanley. “Long-Term Contracts, Rational Expectations, and the
Optimal Money Supply Rule.” Journal of Political Economy 85 (February
1977): 191—205.
Fisher, Irving. “The Business Cycle Largely a ‘Dance of the Dollar.’ “JournaloftheAmerican StatisticalAssociation 18 (December1923): 1024—28.
Fisher, Irving. “Our Unstable Dollar and the So-Called Business Cycle.”
Journal of the American Statistical Association 20 (June 1925): 179—202.
Fisher, Irving. “The Debt-Deflation Theory of the Great Depression.”
Econometrica 1 (October 1933): 337—57.
Fisher, Irving. Stable Money. New York: Adelphi Co., 1934.
Hawtrey, R. G. Monetary Reconstruction. London: Longmans, Green and
Co., 1923.
J. M. A Tract on Monetary Reform. New York: Harcourt Brace and
Co., 1923.
McCloskey, Donald N., and Zecher, J. Richard. “The Success of Purchasing
Power Parity: Historical Evidence and Its Implications for Macroeconomics.” In A Retrospective on the Classical Gold Standard, 1821—1931, pp.
121—70, Edited by M. D. Bordo and A. J. Schwartz. Chicago: University of
Chicago Press, 1984.
Pigou, A. C. Industrial Fluctuations. London: Macmillan and Co., Ltd., 1927.
Robertson, D. H. Money. New York: Harcourt Brace and Co., 1922.
Weinstein, Michael M. Recovery and Redistribution under the NIRA.
Amsterdam: North Holland, 1980.
Keynes,
727